Sakana AIEvoLLM-JP

EvoLLM-JP-v1-10B

This experimental 10B release applies evolutionary merging to Japanese language and mathematics models. Its published size differs from each source checkpoint.

Research release

The merged architecture needs its published loader

The official example loads the model with custom model code enabled. The 10B checkpoint is a distinct merged release, rather than a quantization of the 7B variant.EvoLLM-JP-v1-10B official model card

Working with this release

Practical guidance. Review the model code before execution. Measure memory with the chosen precision and generation length instead of treating the name as a memory requirement.

Specifications & access

EvoLLM-JP-v1-10B specifications
DetailValue and conditions
StatusResearch release
AccessDownloadable
Model identifiersSakanaAI/EvoLLM-JP-v1-10B
AncestryShisa Gamma 7B v1, WizardMath 7B V1.1 and Abel 7B 002.EvoLLM-JP-v1-10B official model card
TermsMicrosoft Research License Terms; described as a research prototype.EvoLLM-JP-v1-10B official model card

Related models

  • EvoLLM-JP-v1-7B

    Sakana AI ยท EvoLLM-JP

    A Japanese language-model merge that combines language and mathematics skills.

    Launch date not confirmedResearch releaseDownloadable